Hands-On Machine Learning for Algorithmic Trading

Training the model

We compile the model to use the Adam optimizer (see Chapter 16, Deep Learning) to minimize the MSE between the input data and the reproduction achieved by the autoencoder. To ensure that the autoencoder learns to reproduce the input, we train the model using the same input and output data:

autoencoder.compile(optimizer='adam', loss='mse')autoencoder.fit(x=X_train_scaled, y=X_train_scaled,                epochs=100, batch_size=32,                shuffle=True, validation_split=.1,                callbacks=[tb_callback, early_stopping, checkpointer])